Multi Objective Genetic Algorithm for Training Three Term Backpropagation Network

Multi Objective Evolutionary Algorithms has been applied for learning problem in Artificial Neural Networks to improve the generalization of the training and testing unseen data. This paper proposes the simultaneous optimization method for training Three Term Back Propagation Network (TTBPN) learning using Multi Objective Genetic Algorithm. The Non-dominated Sorting Genetic Algorithm II is applied to optimize the TTBPN structure by simultaneously reducing the error and complexity in terms of number of hidden nodes of the network for better accuracy in classification problem. This methodology is applied in two kinds of multi classes data set obtained from the University of California at Irvine repository. The results obtained for training and testing on the datasets illustrate less network error and better classification accuracy, besides having simple architecture for the TTBPN.